In order to improve the prediction accuracy of rural road pavement performance, for the practical problems of many factors affecting the pavement performance of rural roads, few years of data accumulation and low prediction accuracy of traditional grey model, this paper adopts the grey model after entropy assignment processing to predict the pavement performance of rural roads, and amends the prediction results of the grey model with the Particle Swarm Algorithm and Markov Model to form an improved grey model. Combined with the test data of pavement performance of cement pavement of rural roads in Hunan Province in recent years, under the premise of analyzing the influencing factors of pavement performance, the cement pavement of rural roads is classified into 6 categories with the type of grass-roots level, the thickness of surface layer, and the type of drainage facilities, and the improved grey model is used to predict the performance of pavement of the 6 categories of rural roads. The results show that relative to the traditional grey prediction model, the model improved by Markov model and particle swarm algorithm significantly reduces the absolute value of relative error and root mean square error between the predicted value and the original value of the A—F class roads, and the average reduction of the absolute value of the relative error is 50.17%, and the average reduction of the root mean square error is 0.73, which improves the prediction accuracy. It shows that the improved grey model model can accurately predict the pavement performance of rural roads, and the prediction results can reflect the changes of pavement performance of Class A-F roads in the predicted years.